Should I vibe code
Desktop and web diagramming with templates for technical and business visuals
Boxes and arrows are a weekend. The stencil library, the Visio import and the print layout are the product.
?
Their verdict, the Annual (Individuals) price and the build-time estimate come from their entry, MIT-licensed. Checked 2026-08-04.
?
Our verdict, the regret score and everything below it. Editorial and unsponsored — nobody can pay to be moved.
The honest answer
why the verdict is what it is
Nothing here can hurt anyone, which is why this is a plain DEMO ONLY rather than something sharper: the worst outcome is a folder of files only your program can open. An agent will hand you an infinite canvas with snapping connectors and a shape palette in an afternoon, and it will look startlingly close to the screenshots. The gap is not the canvas — it is everything Edraw has been accumulating for two decades. Stencil sets for network gear, electrical symbols, floor-plan furniture, org-chart layouts. Visio round-tripping that survives a client sending back a .vsdx. Print and page setup, which is unglamorous and enormous and the actual reason people still buy desktop diagramming software. Connector routing that stays sensible when you drag a box through three other boxes. You can build the demo in a weekend. You cannot build the library, and the library is what you were paying for.
What actually breaks
not "if". the specific failures.
- Your own file format, two refactors later, when the loader can no longer read the diagrams you drew in month one
- Connector routing — trivial for two boxes, an open research problem once a diagram has forty and you drag one through the middle
- Export fidelity: the SVG looks right in the browser and wrong in the PDF, and both are wrong in Word
- Text layout inside shapes, where every homemade canvas goes to die — wrapping, autosize, and fonts the viewer does not have
- Undo, the moment you add grouping, multi-select, and a resize handle that also moves anchor points
- Visio interop, which people assume works because the tool is a diagramming tool, and which is a specification you did not read
- Print: page breaks, scale-to-fit and margins, all needed on the day someone actually presents the diagram
Is that you?
the verdict is a default, not a law
- You need one diagram type your team draws constantly, not a general canvas
- The output is an image you paste elsewhere, and the source of truth is text or a script
- It is local-only and the files land in a directory your existing backup already covers
- You would genuinely rather maintain a small tool that fits your workflow than learn a big one that does not
- Anyone outside you needs to open the files
- The diagrams have to outlive your interest in maintaining the reader
- Someone will eventually want a Visio or PDF export that a client accepts without comment
- You are replacing a paid tool mainly to save the money — this one is cheap enough that money is not the reason
If you build it anyway
the checklist, then the prompt that enforces it
- Do not invent a file format. Serialise to something already specified — SVG with your own metadata, or a documented JSON schema carrying a version field on every document from the first save.
- Write the importer for your own format before the exporter is finished, and keep a fixture file from day one that every build still has to open.
- Pick a canvas library rather than a raw canvas element. tldraw, Excalidraw's editor package, JointJS and Konva have already solved selection, transform handles and hit-testing.
- Autosave to a real file on disk, atomically, and keep the previous version. A canvas app that eats an hour of work loses your trust permanently.
- Decide early whether text lives inside shapes or beside them; retrofitting in-shape layout means rewriting hit-testing, resize and export at the same time.
- Export to SVG first and derive PNG and PDF from it, so there is one rendering path to get wrong instead of three.
- Scope the shape library explicitly. Ten shapes you drew yourself, not a promise of a stencil catalogue you will never finish.
I am building a diagramming canvas. Nothing in it is dangerous, so the whole job is making sure the files still open in a year. Order the work that way and push back if I ask for features before persistence. 1. First, the document format. A versioned JSON schema, a version field written on every save from the first commit, and a loader that rejects unknown versions loudly instead of silently dropping shapes. 2. Then a round-trip test: build a fixture document containing every element type, save it, load it, assert deep equality. That test runs on every change and is the reason this tool is still usable next year. 3. Then saving. Temp file, fsync, rename, and keep the previous version alongside. Never truncate the open document. 4. Only then the canvas. Do not hand-roll selection, transform handles or hit-testing — use tldraw, the Excalidraw editor package, Konva or JointJS, and tell me which you chose and why. 5. Do not write a connector routing engine. Use the library's, or a simple orthogonal router, and accept that dragging a box through a cluster will look bad. Say so rather than iterating on it for a week. 6. Export path: render to SVG, then derive PNG and PDF from that SVG. One rendering path, not three. Show me the same document in all three before any new shape type is added. 7. Fonts: embed or subset whatever the export needs. A diagram that renders correctly on my machine and wrong on the client's is the failure mode here. 8. Text inside shapes is a design decision, not a feature. Settle it before anything else is built, because retrofitting it rewrites hit-testing, resize and export together. 9. Explicitly out of scope, and put this in the README: Visio import/export, real-time collaboration, print and page setup, and any stencil library beyond the ten shapes I actually draw. If I ask for Visio interop, tell me that is exactly where the $69 a year goes.
That one keeps you out of trouble. For the prompt that actually builds it, canivibecodeit.com has one.
their build prompt ↗Or don’t build it
the boring option, and the way back out
The moment a diagram has to leave your machine in a shape somebody else can edit. Sixty-nine dollars a year, or a one-off $198 for the perpetual desktop licence, buys two decades of stencils, a Visio importer and a print dialog that works. And if the appeal was the price rather than the ownership, draw.io is free, runs offline, opens .vsdx and costs nothing at all.
$5.75/mo is cheaper than your weekend.
Export every document to SVG on a schedule rather than on demand. SVG is the one format that still opens in twenty years without your code, and it survives your loader rotting. Keep the JSON alongside for anything you might want to edit again, plus a written note of what the schema fields mean — a versioned format with no documentation is a format only your program understands, which is the same as no format at all.
The closest free equivalent: shape libraries, Visio import and offline desktop builds, still actively developed.
MIT canvas you can embed as a package rather than rebuild, with an approachable interaction model.
Questions
Isn't a diagram editor just shapes and lines?
Shapes and lines are the first afternoon. What takes years is everything that happens when there are a hundred of them: routing that avoids other shapes, text that wraps inside a box being resized, selection that behaves when items are grouped, and an export that looks the same on paper as on screen. None of that is hard to start and all of it is hard to finish.
Why only DEMO ONLY when it is genuinely hard to build?
Because the verdict is about consequences, not difficulty. A half-finished diagram tool costs you evenings and leaves files nobody else can open — bad for you, and for nobody else. Difficulty lives on the build-ease axis instead, and this entry scores low there precisely because the work is real.
What is the honest smaller version worth building?
A generator for the one diagram type you draw constantly, with the source kept as text. Mermaid or D2 does the rendering, your thin layer does the domain. You end up with something better than EdrawMax for that one job and you never touch a selection handle or a print dialog.
Every week, someone ships something they shouldn’t have.
New verdicts, the worst thing that landed in the trap, and the occasional incident report. No other email, ever.
Boxes and arrows is an afternoon. Arrows that stay attached when you move the box is a career.
Graph layout has been free since Graphviz. What you can't download is a symbol set an electrician trusts.
Multiplayer is not a feature you bolt on later. It is a document model, and you picked one on day one.
last reviewed 2026-08-05 · verdict is editorial and unsponsored · shared entry data from canivibecodeit under MIT · not legal advice